More than a hundred storage vendors; over a decade, only fewer than 10 survived.

This industry went through a massacre you didn’t notice.
The few that lived on now hold the same asset the whole world is fighting over: AI memory.
Everyone watches GPUs, but GPUs are the brain—without memory, it’s scrap.
Next to every AI chip, you need huge amounts of DRAM, NAND, and SSD.
In data centers, storage procurement spending accounts for 15-20% of IT equipment spend, and as AI models get bigger, this share keeps rising.
There are currently only three and a half global DRAM manufacturers left: Samsung, SK hynix, Micron, plus a trailing-margin player, ChangXin.
For NAND and AI-grade SSDs, the key players are about 10 firms when you combine the US, Korea, Japan, and China.
Can other companies enter? Even semiconductor giants like TSMC and Intel can’t do storage.
This is heavy asset, heaviest asset of all—requiring continuous national-level capital to pour in for years to even get started. It’s not a business you can run after raising a round of money.
Going from more than 100 to fewer than 10 isn’t because others didn’t want to do it; it’s because the barriers are so high that only national teams can afford to play—supply is highly concentrated, while demand is accelerating its expansion.
In any industry, this kind of structure means one thing: pricing power.
So you understand why Japan is asking Micron to expand production, and the US is asking SK hynix to expand its investment in the US.
It’s not politeness—it’s genuine scarcity, so severe they have to pull you in to build a factory.
This round of adjustment is indeed big: deals got crowded and hedge funds precisely targeted, with the sector down by around 13% in July—but looking from the mid-term perspective, it’s about this much.
After SK hynix and Micron start mass production of HBM4 next year, the logic will fundamentally change.
Customized services, multi-year contracts, customer volume-locking and price-locking—these are not features of cyclical stocks; they are features of platform-type companies, and PE needs to be repriced.
In the short term, focus on risk control; in the long term, storage certainty is too strong.
Supply is just a few firms; demand keeps rising year after year; customers are lining up with multi-year contracts.